Somewhere in a company's sustainability report there is a line saying it went carbon neutral last year. Behind that line sits a certificate. Behind the certificate sits a forest. And behind the forest sits a guess about what would have happened to those trees if nobody had paid anyone anything.

For a long time nobody checked the guess. Then people did.

In October last year a team of researchers published a study in Science that went through 52 forest protection projects across 12 tropical countries, all certified under the same standard that dominates the voluntary carbon market. They rebuilt, from satellite data, what the deforestation in each area would most likely have looked like without the project, and compared it to what actually happened.

Only 19 percent of the projects hit the emissions targets they had reported. About 35 percent had told the registry their forests were under a threat level that the data did not support. And when the researchers looked at the credits themselves, the picture was worse than the project scorecards suggested. Up to 228 million credits had been issued from those projects by the end of 2022. Around 127 million of them had already been used by companies and individuals to cancel out their own emissions. Roughly 35 million looked like they represented real avoided deforestation.

That is about one credit in eight.

This is not one rogue study. A meta analysis published in Nature Communications in 2024 pulled together 14 rigorous studies covering 2,346 crediting projects, close to a billion tonnes of claimed reductions, and found that fewer than 16 percent of the credits examined stood up. For clean cookstove projects it was 11 percent. For avoided deforestation, 25 percent.

Why does this keep happening?

It is tempting to read all this as fraud. Mostly it is not. It is a measurement problem that got sold as a product.

A forest protection credit is payment for something that did not happen. The trees are still standing, so the value of the credit depends entirely on an argument that they were about to come down and did not, because of the money. That argument is called a baseline. The developer builds it, a registry approves it, and an auditor checks that the paperwork is consistent.

Notice who has an interest in the baseline being high. A more threatened forest generates more credits. Nobody has to lie for the number to drift upwards. A slightly pessimistic assumption here, a comparison region chosen a little carelessly there, repeated across a hundred projects, and the market ends up holding paper that a satellite can quietly contradict a decade later.

The buyers found this out the same way everyone else did, by reading it in the press. Some of them are now facing questions about claims they made in good faith on the basis of certificates they paid for.

What you can weigh

Here is the part of my job I find quietly reassuring.

At a palm processing facility in Cross River State, Nigeria, Relief Earth put 255.77 tonnes of palm nut shells through a pyrolysis reactor. What came out was 83.33 tonnes of biochar. Both numbers came off a scale. The biochar was applied to land, the carbon in it is in a form that does not readily rot back into the air, and 190 verified removal credits were issued and bought by Salesforce and Klarna through Milkywire. They were the first verified industrial biochar credits produced in Nigeria.

There is no story in that sequence about what the shells would otherwise have done. There is no comparison region, no satellite reconstruction, no argument about threat levels. There is a weight going in, a weight coming out, and a lab analysis of what fraction of that weight is carbon. Somebody who has never heard the word additionality can stand next to the pile and understand the product.

That is the whole commercial case, and it explains something that looks strange from outside. In the first quarter of this year, biochar accounted for 93 percent of all durable carbon removal volume contracted worldwide. Buyers are not choosing it because it is exciting. They are choosing it because it is boring enough to survive an audit.

The honest caveat

Biochar is not exempt from scrutiny and anybody who tells you otherwise is selling something. The open questions are real: how long the carbon actually stays put under different soil and climate conditions, how accurately carbon content is measured across different reactors and feedstocks, and whether a supplier is producing what they say at the volumes they claim. Registries are still tightening those rules and they should.

The difference is in the kind of question. Forest credits ask you to accept a claim about an alternative history. Biochar asks you to accept a claim about a measurement. Measurements can be wrong, but they can also be repeated, checked, and corrected. Alternative histories cannot.

For anyone in Africa deciding whether to put land, residue or reputation into a carbon project, that distinction is worth more than any price quote. Ask what physical thing gets weighed, who weighs it, and what happens if the number turns out to be wrong. If the answer is a model rather than a scale, that is not automatically a bad project. It just means you are being asked to trust a story, and you should price that accordingly.